File size: 19,251 Bytes
22087d0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 | import os
import io
import uuid
import re
from typing import Dict, Any
from pathlib import Path
import mimetypes
import chardet
import PyPDF2
import docx
from core.utils.logger import logger
from core.services.supabase import DBConnection
from core.services.llm import make_llm_api_call
async def _invalidate_kb_cache_for_entry(client, entry_id: str) -> None:
"""Invalidate KB context cache for all agents that have this entry assigned."""
try:
from core.cache.runtime_cache import invalidate_kb_context_cache
result = await client.table('agent_knowledge_entry_assignments').select(
'agent_id'
).eq('entry_id', entry_id).execute()
agent_ids = set(row['agent_id'] for row in (result.data or []))
for agent_id in agent_ids:
await invalidate_kb_context_cache(agent_id)
if agent_ids:
logger.debug(f"🗑️ Invalidated KB cache for {len(agent_ids)} agents after entry summary update")
except Exception as e:
logger.warning(f"Failed to invalidate KB cache after entry update: {e}")
class FileProcessor:
SUPPORTED_EXTENSIONS = {'.txt', '.pdf', '.docx'}
MAX_FILE_SIZE = 50 * 1024 * 1024
def __init__(self):
self.db = DBConnection()
def sanitize_filename(self, filename: str) -> str:
"""Sanitize filename for S3 storage - remove/replace invalid characters."""
# Keep the file extension
name, ext = os.path.splitext(filename)
# Replace emojis and other problematic Unicode characters with underscores
name = re.sub(r'[^\w\s\-\.]', '_', name)
# Replace spaces with underscores
name = re.sub(r'\s+', '_', name)
# Remove multiple consecutive underscores
name = re.sub(r'_+', '_', name)
# Remove leading/trailing underscores
name = name.strip('_')
# Ensure name is not empty
if not name:
name = 'file'
return f"{name}{ext}"
def _is_likely_text_file(self, file_content: bytes) -> bool:
"""Check if file content is likely text-based."""
try:
# Try to decode as text
detected = chardet.detect(file_content[:1024]) # Check first 1KB
if detected.get('confidence', 0) > 0.7:
decoded = file_content[:1024].decode(detected.get('encoding', 'utf-8'))
# Check if most characters are printable
printable_ratio = len([c for c in decoded if c.isprintable() or c.isspace()]) / len(decoded)
return printable_ratio > 0.8
except:
pass
return False
async def process_file_fast(
self,
account_id: str,
folder_id: str,
file_content: bytes,
filename: str,
mime_type: str,
background_tasks
) -> Dict[str, Any]:
"""Fast file processing - upload file immediately, generate summary in background."""
import time
start = time.time()
logger.info(f"[PROCESSOR] Starting fast processing for: {filename}")
try:
if len(file_content) > self.MAX_FILE_SIZE:
raise ValueError(f"File too large: {len(file_content)} bytes")
file_extension = Path(filename).suffix.lower()
# Check if it's text-based first
is_text_based = (
mime_type.startswith('text/') or
mime_type in ['application/json', 'application/xml', 'text/xml'] or
self._is_likely_text_file(file_content)
)
# If not text-based, check allowed extensions
if not is_text_based and file_extension not in self.SUPPORTED_EXTENSIONS:
raise ValueError(f"Unsupported file type: {file_extension}")
# Generate unique entry ID
entry_id = str(uuid.uuid4())
# Sanitize filename for S3 storage
sanitized_filename = self.sanitize_filename(filename)
# Upload to S3
s3_path = f"knowledge-base/{folder_id}/{entry_id}/{sanitized_filename}"
client = await self.db.client
t1 = time.time()
await client.storage.from_('file-uploads').upload(
s3_path, file_content, {"content-type": mime_type}
)
logger.info(f"[PROCESSOR] S3 upload took: {time.time() - t1:.2f}s")
# Save to database with placeholder summary
entry_data = {
'entry_id': entry_id,
'folder_id': folder_id,
'account_id': account_id,
'filename': filename,
'file_path': s3_path,
'file_size': len(file_content),
'mime_type': mime_type,
'summary': 'Processing...',
'is_active': True
}
t2 = time.time()
await client.table('knowledge_base_entries').insert(entry_data).execute()
logger.info(f"[PROCESSOR] DB insert took: {time.time() - t2:.2f}s")
# Schedule background summary generation
t3 = time.time()
background_tasks.add_task(
self._generate_and_update_summary,
entry_id,
file_content,
filename,
mime_type
)
logger.info(f"[PROCESSOR] Background task scheduled in: {time.time() - t3:.2f}s")
logger.info(f"[PROCESSOR] Total fast processing time: {time.time() - start:.2f}s")
return {
'success': True,
'entry_id': entry_id,
'filename': filename,
'processing': True
}
except Exception as e:
logger.error(f"Error processing file {filename}: {str(e)}")
return {'success': False, 'error': str(e)}
async def _generate_and_update_summary(
self,
entry_id: str,
file_content: bytes,
filename: str,
mime_type: str
):
"""Background task to generate and update file summary."""
try:
# Extract content
content = self._extract_content(file_content, filename, mime_type)
if not content:
content = f"File: {filename} ({len(file_content)} bytes, {mime_type})"
# Generate summary
summary = await self._generate_summary(content, filename)
# Update database
client = await self.db.client
await client.table('knowledge_base_entries').update({
'summary': summary
}).eq('entry_id', entry_id).execute()
await _invalidate_kb_cache_for_entry(client, entry_id)
logger.info(f"Successfully generated summary for entry {entry_id}")
except Exception as e:
logger.error(f"Error generating summary for entry {entry_id}: {str(e)}")
# Update with error message
try:
client = await self.db.client
await client.table('knowledge_base_entries').update({
'summary': f"Error generating summary: {str(e)}"
}).eq('entry_id', entry_id).execute()
await _invalidate_kb_cache_for_entry(client, entry_id)
except Exception:
pass
async def process_file(
self,
account_id: str,
folder_id: str,
file_content: bytes,
filename: str,
mime_type: str
) -> Dict[str, Any]:
"""Legacy synchronous file processing - kept for backward compatibility."""
try:
if len(file_content) > self.MAX_FILE_SIZE:
raise ValueError(f"File too large: {len(file_content)} bytes")
file_extension = Path(filename).suffix.lower()
# Check if it's text-based first
is_text_based = (
mime_type.startswith('text/') or
mime_type in ['application/json', 'application/xml', 'text/xml'] or
self._is_likely_text_file(file_content)
)
# If not text-based, check allowed extensions
if not is_text_based and file_extension not in self.SUPPORTED_EXTENSIONS:
raise ValueError(f"Unsupported file type: {file_extension}")
# Generate unique entry ID
entry_id = str(uuid.uuid4())
# Sanitize filename for S3 storage
sanitized_filename = self.sanitize_filename(filename)
# Upload to S3
s3_path = f"knowledge-base/{folder_id}/{entry_id}/{sanitized_filename}"
client = await self.db.client
await client.storage.from_('file-uploads').upload(
s3_path, file_content, {"content-type": mime_type}
)
# Extract content for summary
content = self._extract_content(file_content, filename, mime_type)
if not content:
# If no content could be extracted, create a basic file info summary
content = f"File: {filename} ({len(file_content)} bytes, {mime_type})"
# Generate LLM summary
summary = await self._generate_summary(content, filename)
# Save to database
entry_data = {
'entry_id': entry_id,
'folder_id': folder_id,
'account_id': account_id,
'filename': filename,
'file_path': s3_path,
'file_size': len(file_content),
'mime_type': mime_type,
'summary': summary,
'is_active': True
}
result = await client.table('knowledge_base_entries').insert(entry_data).execute()
return {
'success': True,
'entry_id': entry_id,
'filename': filename,
'summary_length': len(summary)
}
except Exception as e:
logger.error(f"Error processing file {filename}: {str(e)}")
return {'success': False, 'error': str(e)}
async def _generate_summary(self, content: str, filename: str) -> str:
try:
models = [
("openrouter/google/gemini-2.5-flash-lite", 1_000_000),
("gpt-5-mini", 400_000)
]
estimated_tokens = len(content) // 4
for model_name, context_limit in models:
try:
usable_context = context_limit - 1000
if estimated_tokens <= usable_context:
processed_content = content
else:
processed_content = self._smart_chunk_content(content, usable_context * 4) # Convert back to chars
prompt = f"""Analyze this file and create a concise, actionable summary for an AI agent's knowledge base.
File: {filename}
Content: {processed_content}
Generate a 2-3 sentence summary that captures:
1. What this file contains
2. Key information or purpose
3. When this knowledge would be useful
Keep it under 200 words and make it actionable for context injection."""
messages = [{"role": "user", "content": prompt}]
response = await make_llm_api_call(
messages=messages,
model_name=model_name,
temperature=0.1,
max_tokens=300,
stream=False,
)
summary = ""
if isinstance(response, dict):
summary = (
(response.get("choices") or [{}])[0]
.get("message", {})
.get("content", "")
.strip()
)
elif hasattr(response, "choices") and response.choices:
summary = (response.choices[0].message.content or "").strip()
else:
summary = str(response).strip()
if summary:
logger.info(f"Summary generated successfully using {model_name}")
return summary
except Exception as e:
logger.warning(f"Model {model_name} failed: {str(e)}")
continue
# All models failed - high-reliability fallback
logger.error("All LLM models failed, using intelligent fallback")
return self._create_fallback_summary(content, filename)
except Exception as e:
logger.error(f"Error generating summary: {str(e)}")
return self._create_fallback_summary(content, filename)
def _smart_chunk_content(self, content: str, max_chars: int) -> str:
"""Intelligently chunk content to fit context limits."""
if len(content) <= max_chars:
return content
# Strategy: Take beginning + end with priority sections
quarter = max_chars // 4
# Always include beginning (most important)
beginning = content[:quarter * 2]
# Include end for conclusion/summary
ending = content[-quarter:]
# Try to find important middle sections (headers, key terms)
middle_section = ""
remaining_chars = max_chars - len(beginning) - len(ending) - 100 # Buffer
if remaining_chars > 0:
middle_start = len(beginning)
middle_end = len(content) - quarter
middle_content = content[middle_start:middle_end]
# Look for important sections (lines with caps, numbers, bullets)
lines = middle_content.split('\n')
important_lines = []
current_length = 0
for line in lines:
line_stripped = line.strip()
if not line_stripped:
continue
# Prioritize lines that look important
is_important = (
line_stripped.isupper() or # Headers
line_stripped.startswith(('•', '-', '*', '1.', '2.')) or # Lists
any(keyword in line_stripped.lower() for keyword in ['summary', 'conclusion', 'important', 'key', 'main']) or
len(line_stripped) < 100 # Short lines often important
)
if is_important and current_length + len(line) < remaining_chars:
important_lines.append(line)
current_length += len(line)
if important_lines:
middle_section = '\n'.join(important_lines)
# Combine sections
if middle_section:
return f"{beginning}\n\n[KEY SECTIONS]\n{middle_section}\n\n[ENDING]\n{ending}"
else:
return f"{beginning}\n\n[...content truncated...]\n\n{ending}"
def _create_fallback_summary(self, content: str, filename: str) -> str:
"""Create intelligent fallback summary when LLM fails."""
# Extract first meaningful portion
preview = content[:2000].strip()
# Basic content analysis
lines = content.split('\n')
non_empty_lines = [line.strip() for line in lines if line.strip()]
# Try to identify content type
content_type = "document"
if filename.endswith('.py'):
content_type = "Python code"
elif filename.endswith('.js'):
content_type = "JavaScript code"
elif filename.endswith('.md'):
content_type = "Markdown document"
elif filename.endswith('.txt'):
content_type = "text document"
elif filename.endswith('.csv'):
content_type = "CSV data file"
# Generate intelligent fallback
return f"This {content_type} '{filename}' contains {len(content):,} characters across {len(non_empty_lines)} lines. Preview: {preview[:200]}{'...' if len(preview) > 200 else ''} This file would be useful for understanding the specific content and context it provides."
def _extract_content(self, file_content: bytes, filename: str, mime_type: str) -> str:
"""Extract text content from file bytes."""
file_extension = Path(filename).suffix.lower()
try:
# Handle text-based files (including JSON, XML, CSV, etc.)
if (file_extension in ['.txt', '.json', '.xml', '.csv', '.yml', '.yaml', '.md', '.log', '.ini', '.cfg', '.conf']
or mime_type.startswith('text/')
or mime_type in ['application/json', 'application/xml', 'text/xml']):
detected = chardet.detect(file_content)
encoding = detected.get('encoding', 'utf-8')
try:
return file_content.decode(encoding)
except UnicodeDecodeError:
return file_content.decode('utf-8', errors='replace')
elif file_extension == '.pdf':
pdf_reader = PyPDF2.PdfReader(io.BytesIO(file_content))
return '\n\n'.join(page.extract_text() for page in pdf_reader.pages)
elif file_extension == '.docx':
doc = docx.Document(io.BytesIO(file_content))
return '\n'.join(paragraph.text for paragraph in doc.paragraphs)
# For any other file type, try to decode as text (fallback)
else:
try:
detected = chardet.detect(file_content)
encoding = detected.get('encoding', 'utf-8')
content = file_content.decode(encoding)
# Only return if it seems to be mostly text content
if len([c for c in content[:1000] if c.isprintable() or c.isspace()]) > 800:
return content
except:
pass
# If we can't extract text content, return a placeholder
return f"[Binary file: {filename}] - Content cannot be extracted as text, but file is stored and available for download."
except Exception as e:
logger.error(f"Error extracting content from {filename}: {str(e)}")
return f"[Error extracting content from {filename}] - File is stored but content extraction failed: {str(e)}" |